xclele/yolov10
1
0
Fork 0
mirror of https://github.com/THU-MIG/yolov10.git synced 2025-05-23 05:24:22 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

add yolov10

Browse Source
This commit is contained in:
wa22 2024-05-23 05:57:37 +00:00
parent 5be2ffbd13
commit 8d9f45a955
31 changed files with 4947 additions and 599 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

342
README.md
View File

@ -1,295 +1,85 @@
<div align="center"> # [YOLOv10: Real-Time End-to-End Object Detection]()
<p>
<a href="https://www.ultralytics.com/blog/ultralytics-yolov8-turns-one-a-year-of-breakthroughs-and-innovations" target="_blank">
<img width="100%" src="https://raw.githubusercontent.com/ultralytics/assets/main/im/banner-yolo-vision-2023.png" alt="YOLO Vision banner"></a> Official PyTorch implementation of **YOLOv10**.
<p align="center">
<img src="figures/latency.svg" width=48%>
<img src="figures/params.svg" width=48%> <br>
Comparisons with others in terms of latency-accuracy (left) and size-accuracy (right) trade-offs.
</p> </p>
[中文](https://docs.ultralytics.com/zh/) | [한국어](https://docs.ultralytics.com/ko/) | [日本語](https://docs.ultralytics.com/ja/) | [Русский](https://docs.ultralytics.com/ru/) | [Deutsch](https://docs.ultralytics.com/de/) | [Français](https://docs.ultralytics.com/fr/) | [Español](https://docs.ultralytics.com/es/) | [Português](https://docs.ultralytics.com/pt/) | [हिन्दी](https://docs.ultralytics.com/hi/) | [العربية](https://docs.ultralytics.com/ar/) <br> [YOLOv10: Real-Time End-to-End Object Detection]().\
Ao Wang, Hui Chen, Lihao Liu, Kai Chen, Zijia Lin, Jungong Han, and Guiguang Ding\
[[`arXiv`]()]
<div> <details>
<a href="https://github.com/ultralytics/ultralytics/actions/workflows/ci.yaml"><img src="https://github.com/ultralytics/ultralytics/actions/workflows/ci.yaml/badge.svg" alt="Ultralytics CI"></a> <summary>
<a href="https://codecov.io/github/ultralytics/ultralytics"><img src="https://codecov.io/github/ultralytics/ultralytics/branch/main/graph/badge.svg?token=HHW7IIVFVY" alt="Ultralytics Code Coverage"></a> <font size="+1">Abstract</font>
<a href="https://zenodo.org/badge/latestdoi/264818686"><img src="https://zenodo.org/badge/264818686.svg" alt="YOLOv8 Citation"></a> </summary>
<a href="https://hub.docker.com/r/ultralytics/ultralytics"><img src="https://img.shields.io/docker/pulls/ultralytics/ultralytics?logo=docker" alt="Docker Pulls"></a> Over the past years, YOLOs have emerged as the predominant paradigm in the field of real-time object detection owing to their effective balance between computational cost and detection performance. Researchers have explored the architectural designs, optimization objectives, data augmentation strategies, and others for YOLOs, achieving notable progress. However, the reliance on the non-maximum suppression (NMS) for post-processing hampers the end-to-end deployment of YOLOs and adversely impacts the inference latency. Besides, the design of various components in YOLOs lacks the comprehensive and thorough inspection, resulting in noticeable computational redundancy and limiting the model's capability. It renders the suboptimal efficiency, along with considerable potential for performance improvements. In this work, we aim to further advance the performance-efficiency boundary of YOLOs from both the post-processing and the model architecture. To this end, we first present the consistent dual assignments for NMS-free training of YOLOs, which brings the competitive performance and low inference latency simultaneously. Moreover, we introduce the holistic efficiency-accuracy driven model design strategy for YOLOs. We comprehensively optimize various components of YOLOs from both the efficiency and accuracy perspectives, which greatly reduces the computational overhead and enhances the capability. The outcome of our effort is a new generation of YOLO series for real-time end-to-end object detection, dubbed YOLOv10. Extensive experiments show that YOLOv10 achieves the state-of-the-art performance and efficiency across various model scales. For example, our YOLOv10-S is 1.8$\times$ faster than RT-DETR-R18 under the similar AP on COCO, meanwhile enjoying 2.8$\times$ smaller number of parameters and FLOPs. Compared with YOLOv9-C, YOLOv10-B has 46\% less latency and 25\% fewer parameters for the same performance.
<a href="https://ultralytics.com/discord"><img alt="Discord" src="https://img.shields.io/discord/1089800235347353640?logo=discord&logoColor=white&label=Discord&color=blue"></a> </details>
<br>
<a href="https://console.paperspace.com/github/ultralytics/ultralytics"><img src="https://assets.paperspace.io/img/gradient-badge.svg" alt="Run on Gradient"></a>
<a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/tutorial.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a>
<a href="https://www.kaggle.com/ultralytics/yolov8"><img src="https://kaggle.com/static/images/open-in-kaggle.svg" alt="Open In Kaggle"></a>
</div>
<br>
[Ultralytics](https://ultralytics.com) [YOLOv8](https://github.com/ultralytics/ultralytics) is a cutting-edge, state-of-the-art (SOTA) model that builds upon the success of previous YOLO versions and introduces new features and improvements to further boost performance and flexibility. YOLOv8 is designed to be fast, accurate, and easy to use, making it an excellent choice for a wide range of object detection and tracking, instance segmentation, image classification and pose estimation tasks. ## Performance
COCO
| Model | Test Size | #Params | FLOPs | AP<sup>val</sup> | Latency |
|:---------------|:----:|:---:|:--:|:--:|:--:|
| [YOLOv10-N](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10n.pt) | 640 | 2.3M | 6.7G | 38.5% | 1.84ms |
| [YOLOv10-S](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10s.pt) | 640 | 7.2M | 21.6G | 46.3% | 2.49ms |
| [YOLOv10-M](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10m.pt) | 640 | 15.4M | 59.1G | 51.1% | 4.74ms |
| [YOLOv10-B](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10b.pt) | 640 | 19.1M | 92.0G | 52.5% | 5.74ms |
| [YOLOv10-L](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10l.pt) | 640 | 24.4M | 120.3G | 53.2% | 7.28ms |
| [YOLOv10-X](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10x.pt) | 640 | 29.5M | 160.4G | 54.4% | 10.70ms |
We hope that the resources here will help you get the most out of YOLOv8. Please browse the YOLOv8 <a href="https://docs.ultralytics.com/">Docs</a> for details, raise an issue on <a href="https://github.com/ultralytics/ultralytics/issues/new/choose">GitHub</a> for support, and join our <a href="https://ultralytics.com/discord">Discord</a> community for questions and discussions! ## Installation
`conda` virtual environment is recommended.
To request an Enterprise License please complete the form at [Ultralytics Licensing](https://ultralytics.com/license). ```
conda create -n yolov10 python=3.9
<img width="100%" src="https://raw.githubusercontent.com/ultralytics/assets/main/yolov8/yolo-comparison-plots.png" alt="YOLOv8 performance plots"></a> conda activate yolov10
pip install -r requirements.txt
<div align="center"> pip install -e .
<a href="https://github.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-github.png" width="2%" alt="Ultralytics GitHub"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://www.linkedin.com/company/ultralytics/"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-linkedin.png" width="2%" alt="Ultralytics LinkedIn"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://twitter.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-twitter.png" width="2%" alt="Ultralytics Twitter"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://youtube.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-youtube.png" width="2%" alt="Ultralytics YouTube"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://www.tiktok.com/@ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-tiktok.png" width="2%" alt="Ultralytics TikTok"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://www.instagram.com/ultralytics/"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-instagram.png" width="2%" alt="Ultralytics Instagram"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://ultralytics.com/discord"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-discord.png" width="2%" alt="Ultralytics Discord"></a>
</div>
</div>
## <div align="center">Documentation</div>
See below for a quickstart installation and usage example, and see the [YOLOv8 Docs](https://docs.ultralytics.com) for full documentation on training, validation, prediction and deployment.
<details open>
<summary>Install</summary>
Pip install the ultralytics package including all [requirements](https://github.com/ultralytics/ultralytics/blob/main/pyproject.toml) in a [**Python>=3.8**](https://www.python.org/) environment with [**PyTorch>=1.8**](https://pytorch.org/get-started/locally/).
[![PyPI version](https://badge.fury.io/py/ultralytics.svg)](https://badge.fury.io/py/ultralytics) [![Downloads](https://static.pepy.tech/badge/ultralytics)](https://pepy.tech/project/ultralytics)
```bash
pip install ultralytics
``` ```
For alternative installation methods including [Conda](https://anaconda.org/conda-forge/ultralytics), [Docker](https://hub.docker.com/r/ultralytics/ultralytics), and Git, please refer to the [Quickstart Guide](https://docs.ultralytics.com/quickstart). ## Validation
[`yolov10n.pt`](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10n.pt) [`yolov10s.pt`](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10s.pt) [`yolov10m.pt`](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10m.pt) [`yolov10b.pt`](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10b.pt) [`yolov10l.pt`](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10l.pt) [`yolov10x.pt`](https://github.com/jameslahm/yolov10/releases/download/v1.0/yolov10x.pt)
</details> ```
yolo val model=yolov10n/s/m/b/l/x.pt data=coco.yaml batch=256
<details open>
<summary>Usage</summary>
### CLI
YOLOv8 may be used directly in the Command Line Interface (CLI) with a `yolo` command:
```bash
yolo predict model=yolov8n.pt source='https://ultralytics.com/images/bus.jpg'
``` ```
`yolo` can be used for a variety of tasks and modes and accepts additional arguments, i.e. `imgsz=640`. See the YOLOv8 [CLI Docs](https://docs.ultralytics.com/usage/cli) for examples. ## Training
```
### Python yolo detect train data=coco.yaml model=yolov10n/s/m/b/l/x.yaml epochs=500 batch=256 imgsz=640 device=0,1,2,3,4,5,6,7
YOLOv8 may also be used directly in a Python environment, and accepts the same [arguments](https://docs.ultralytics.com/usage/cfg/) as in the CLI example above:
```python
from ultralytics import YOLO
# Load a model
model = YOLO("yolov8n.yaml") # build a new model from scratch
model = YOLO("yolov8n.pt") # load a pretrained model (recommended for training)
# Use the model
model.train(data="coco128.yaml", epochs=3) # train the model
metrics = model.val() # evaluate model performance on the validation set
results = model("https://ultralytics.com/images/bus.jpg") # predict on an image
path = model.export(format="onnx") # export the model to ONNX format
``` ```
See YOLOv8 [Python Docs](https://docs.ultralytics.com/usage/python) for more examples. ## Prediction
```
yolo predict model=yolov10n/s/m/b/l/x.pt
```
</details> ## Export
```
# End-to-End ONNX
yolo export model=yolov10n/s/m/b/l/x.pt format=onnx opset=13 simplify
# Predict with ONNX
yolo predict model=yolov10n/s/m/b/l/x.onnx
### Notebooks # End-to-End TensorRT
yolo export model=yolov10n/s/m/b/l/x.pt format=engine half=True simplify opset=13 workspace=16
# Or
trtexec --onnx=yolov10n/s/m/b/l/x.onnx --saveEngine=yolov10n/s/m/b/l/x.engine --fp16
# Predict with TensorRT
yolo predict model=yolov10n/s/m/b/l/x.engine
```
Ultralytics provides interactive notebooks for YOLOv8, covering training, validation, tracking, and more. Each notebook is paired with a [YouTube](https://youtube.com/ultralytics) tutorial, making it easy to learn and implement advanced YOLOv8 features. ## Acknowledgement
| Docs | Notebook | YouTube | The code base is built with [ultralytics](https://github.com/ultralytics/ultralytics) and [RT-DETR](https://github.com/lyuwenyu/RT-DETR).
| --------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| <a href="https://docs.ultralytics.com/modes/">YOLOv8 Train, Val, Predict and Export Modes</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/tutorial.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> | <a href="https://youtu.be/j8uQc0qB91s"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube Video"></center></a> |
| <a href="https://docs.ultralytics.com/hub/quickstart/">Ultralytics HUB QuickStart</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/hub.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> | <a href="https://youtu.be/lveF9iCMIzc"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube Video"></center></a> |
| <a href="https://docs.ultralytics.com/modes/track/">YOLOv8 Multi-Object Tracking in Videos</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/object_tracking.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> | <a href="https://youtu.be/hHyHmOtmEgs"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube Video"></center></a> |
| <a href="https://docs.ultralytics.com/guides/object-counting/">YOLOv8 Object Counting in Videos</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/object_counting.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> | <a href="https://youtu.be/Ag2e-5_NpS0"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube Video"></center></a> |
| <a href="https://docs.ultralytics.com/guides/heatmaps/">YOLOv8 Heatmaps in Videos</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/heatmaps.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> | <a href="https://youtu.be/4ezde5-nZZw"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube Video"></center></a> |
| <a href="https://docs.ultralytics.com/datasets/explorer/">Ultralytics Datasets Explorer with SQL and OpenAI Integration 🚀 New</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/docs/en/datasets/explorer/explorer.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> | <a href="https://youtu.be/3VryynorQeo"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube Video"></center></a> |
## <div align="center">Models</div> Thanks for the great implementations!
YOLOv8 [Detect](https://docs.ultralytics.com/tasks/detect), [Segment](https://docs.ultralytics.com/tasks/segment) and [Pose](https://docs.ultralytics.com/tasks/pose) models pretrained on the [COCO](https://docs.ultralytics.com/datasets/detect/coco) dataset are available here, as well as YOLOv8 [Classify](https://docs.ultralytics.com/tasks/classify) models pretrained on the [ImageNet](https://docs.ultralytics.com/datasets/classify/imagenet) dataset. [Track](https://docs.ultralytics.com/modes/track) mode is available for all Detect, Segment and Pose models. ## Citation
<img width="1024" src="https://raw.githubusercontent.com/ultralytics/assets/main/im/banner-tasks.png" alt="Ultralytics YOLO supported tasks"> If our code or models help your work, please cite our paper:
```BibTeX
All [Models](https://github.com/ultralytics/ultralytics/tree/main/ultralytics/cfg/models) download automatically from the latest Ultralytics [release](https://github.com/ultralytics/assets/releases) on first use. ```
<details open><summary>Detection (COCO)</summary>
See [Detection Docs](https://docs.ultralytics.com/tasks/detect/) for usage examples with these models trained on [COCO](https://docs.ultralytics.com/datasets/detect/coco/), which include 80 pre-trained classes.
| Model | size<br><sup>(pixels) | mAP<sup>val<br>50-95 | Speed<br><sup>CPU ONNX<br>(ms) | Speed<br><sup>A100 TensorRT<br>(ms) | params<br><sup>(M) | FLOPs<br><sup>(B) |
| ------------------------------------------------------------------------------------ | --------------------- | -------------------- | ------------------------------ | ----------------------------------- | ------------------ | ----------------- |
| [YOLOv8n](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n.pt) | 640 | 37.3 | 80.4 | 0.99 | 3.2 | 8.7 |
| [YOLOv8s](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s.pt) | 640 | 44.9 | 128.4 | 1.20 | 11.2 | 28.6 |
| [YOLOv8m](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m.pt) | 640 | 50.2 | 234.7 | 1.83 | 25.9 | 78.9 |
| [YOLOv8l](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l.pt) | 640 | 52.9 | 375.2 | 2.39 | 43.7 | 165.2 |
| [YOLOv8x](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x.pt) | 640 | 53.9 | 479.1 | 3.53 | 68.2 | 257.8 |
- **mAP<sup>val</sup>** values are for single-model single-scale on [COCO val2017](https://cocodataset.org) dataset. <br>Reproduce by `yolo val detect data=coco.yaml device=0`
- **Speed** averaged over COCO val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. <br>Reproduce by `yolo val detect data=coco.yaml batch=1 device=0|cpu`
</details>
<details><summary>Detection (Open Image V7)</summary>
See [Detection Docs](https://docs.ultralytics.com/tasks/detect/) for usage examples with these models trained on [Open Image V7](https://docs.ultralytics.com/datasets/detect/open-images-v7/), which include 600 pre-trained classes.
| Model | size<br><sup>(pixels) | mAP<sup>val<br>50-95 | Speed<br><sup>CPU ONNX<br>(ms) | Speed<br><sup>A100 TensorRT<br>(ms) | params<br><sup>(M) | FLOPs<br><sup>(B) |
| ----------------------------------------------------------------------------------------- | --------------------- | -------------------- | ------------------------------ | ----------------------------------- | ------------------ | ----------------- |
| [YOLOv8n](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-oiv7.pt) | 640 | 18.4 | 142.4 | 1.21 | 3.5 | 10.5 |
| [YOLOv8s](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-oiv7.pt) | 640 | 27.7 | 183.1 | 1.40 | 11.4 | 29.7 |
| [YOLOv8m](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-oiv7.pt) | 640 | 33.6 | 408.5 | 2.26 | 26.2 | 80.6 |
| [YOLOv8l](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-oiv7.pt) | 640 | 34.9 | 596.9 | 2.43 | 44.1 | 167.4 |
| [YOLOv8x](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-oiv7.pt) | 640 | 36.3 | 860.6 | 3.56 | 68.7 | 260.6 |
- **mAP<sup>val</sup>** values are for single-model single-scale on [Open Image V7](https://docs.ultralytics.com/datasets/detect/open-images-v7/) dataset. <br>Reproduce by `yolo val detect data=open-images-v7.yaml device=0`
- **Speed** averaged over Open Image V7 val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. <br>Reproduce by `yolo val detect data=open-images-v7.yaml batch=1 device=0|cpu`
</details>
<details><summary>Segmentation (COCO)</summary>
See [Segmentation Docs](https://docs.ultralytics.com/tasks/segment/) for usage examples with these models trained on [COCO-Seg](https://docs.ultralytics.com/datasets/segment/coco/), which include 80 pre-trained classes.
| Model | size<br><sup>(pixels) | mAP<sup>box<br>50-95 | mAP<sup>mask<br>50-95 | Speed<br><sup>CPU ONNX<br>(ms) | Speed<br><sup>A100 TensorRT<br>(ms) | params<br><sup>(M) | FLOPs<br><sup>(B) |
| -------------------------------------------------------------------------------------------- | --------------------- | -------------------- | --------------------- | ------------------------------ | ----------------------------------- | ------------------ | ----------------- |
| [YOLOv8n-seg](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-seg.pt) | 640 | 36.7 | 30.5 | 96.1 | 1.21 | 3.4 | 12.6 |
| [YOLOv8s-seg](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-seg.pt) | 640 | 44.6 | 36.8 | 155.7 | 1.47 | 11.8 | 42.6 |
| [YOLOv8m-seg](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-seg.pt) | 640 | 49.9 | 40.8 | 317.0 | 2.18 | 27.3 | 110.2 |
| [YOLOv8l-seg](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-seg.pt) | 640 | 52.3 | 42.6 | 572.4 | 2.79 | 46.0 | 220.5 |
| [YOLOv8x-seg](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-seg.pt) | 640 | 53.4 | 43.4 | 712.1 | 4.02 | 71.8 | 344.1 |
- **mAP<sup>val</sup>** values are for single-model single-scale on [COCO val2017](https://cocodataset.org) dataset. <br>Reproduce by `yolo val segment data=coco-seg.yaml device=0`
- **Speed** averaged over COCO val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. <br>Reproduce by `yolo val segment data=coco-seg.yaml batch=1 device=0|cpu`
</details>
<details><summary>Pose (COCO)</summary>
See [Pose Docs](https://docs.ultralytics.com/tasks/pose/) for usage examples with these models trained on [COCO-Pose](https://docs.ultralytics.com/datasets/pose/coco/), which include 1 pre-trained class, person.
| Model | size<br><sup>(pixels) | mAP<sup>pose<br>50-95 | mAP<sup>pose<br>50 | Speed<br><sup>CPU ONNX<br>(ms) | Speed<br><sup>A100 TensorRT<br>(ms) | params<br><sup>(M) | FLOPs<br><sup>(B) |
| ---------------------------------------------------------------------------------------------------- | --------------------- | --------------------- | ------------------ | ------------------------------ | ----------------------------------- | ------------------ | ----------------- |
| [YOLOv8n-pose](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-pose.pt) | 640 | 50.4 | 80.1 | 131.8 | 1.18 | 3.3 | 9.2 |
| [YOLOv8s-pose](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-pose.pt) | 640 | 60.0 | 86.2 | 233.2 | 1.42 | 11.6 | 30.2 |
| [YOLOv8m-pose](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-pose.pt) | 640 | 65.0 | 88.8 | 456.3 | 2.00 | 26.4 | 81.0 |
| [YOLOv8l-pose](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-pose.pt) | 640 | 67.6 | 90.0 | 784.5 | 2.59 | 44.4 | 168.6 |
| [YOLOv8x-pose](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-pose.pt) | 640 | 69.2 | 90.2 | 1607.1 | 3.73 | 69.4 | 263.2 |
| [YOLOv8x-pose-p6](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-pose-p6.pt) | 1280 | 71.6 | 91.2 | 4088.7 | 10.04 | 99.1 | 1066.4 |
- **mAP<sup>val</sup>** values are for single-model single-scale on [COCO Keypoints val2017](https://cocodataset.org) dataset. <br>Reproduce by `yolo val pose data=coco-pose.yaml device=0`
- **Speed** averaged over COCO val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. <br>Reproduce by `yolo val pose data=coco-pose.yaml batch=1 device=0|cpu`
</details>
<details><summary>OBB (DOTAv1)</summary>
See [OBB Docs](https://docs.ultralytics.com/tasks/obb/) for usage examples with these models trained on [DOTAv1](https://docs.ultralytics.com/datasets/obb/dota-v2/#dota-v10/), which include 15 pre-trained classes.
| Model | size<br><sup>(pixels) | mAP<sup>test<br>50 | Speed<br><sup>CPU ONNX<br>(ms) | Speed<br><sup>A100 TensorRT<br>(ms) | params<br><sup>(M) | FLOPs<br><sup>(B) |
| -------------------------------------------------------------------------------------------- | --------------------- | ------------------ | ------------------------------ | ----------------------------------- | ------------------ | ----------------- |
| [YOLOv8n-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-obb.pt) | 1024 | 78.0 | 204.77 | 3.57 | 3.1 | 23.3 |
| [YOLOv8s-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-obb.pt) | 1024 | 79.5 | 424.88 | 4.07 | 11.4 | 76.3 |
| [YOLOv8m-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-obb.pt) | 1024 | 80.5 | 763.48 | 7.61 | 26.4 | 208.6 |
| [YOLOv8l-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-obb.pt) | 1024 | 80.7 | 1278.42 | 11.83 | 44.5 | 433.8 |
| [YOLOv8x-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-obb.pt) | 1024 | 81.36 | 1759.10 | 13.23 | 69.5 | 676.7 |
- **mAP<sup>test</sup>** values are for single-model multiscale on [DOTAv1](https://captain-whu.github.io/DOTA/index.html) dataset. <br>Reproduce by `yolo val obb data=DOTAv1.yaml device=0 split=test` and submit merged results to [DOTA evaluation](https://captain-whu.github.io/DOTA/evaluation.html).
- **Speed** averaged over DOTAv1 val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. <br>Reproduce by `yolo val obb data=DOTAv1.yaml batch=1 device=0|cpu`
</details>
<details><summary>Classification (ImageNet)</summary>
See [Classification Docs](https://docs.ultralytics.com/tasks/classify/) for usage examples with these models trained on [ImageNet](https://docs.ultralytics.com/datasets/classify/imagenet/), which include 1000 pretrained classes.
| Model | size<br><sup>(pixels) | acc<br><sup>top1 | acc<br><sup>top5 | Speed<br><sup>CPU ONNX<br>(ms) | Speed<br><sup>A100 TensorRT<br>(ms) | params<br><sup>(M) | FLOPs<br><sup>(B) at 640 |
| -------------------------------------------------------------------------------------------- | --------------------- | ---------------- | ---------------- | ------------------------------ | ----------------------------------- | ------------------ | ------------------------ |
| [YOLOv8n-cls](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-cls.pt) | 224 | 69.0 | 88.3 | 12.9 | 0.31 | 2.7 | 4.3 |
| [YOLOv8s-cls](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-cls.pt) | 224 | 73.8 | 91.7 | 23.4 | 0.35 | 6.4 | 13.5 |
| [YOLOv8m-cls](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-cls.pt) | 224 | 76.8 | 93.5 | 85.4 | 0.62 | 17.0 | 42.7 |
| [YOLOv8l-cls](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-cls.pt) | 224 | 76.8 | 93.5 | 163.0 | 0.87 | 37.5 | 99.7 |
| [YOLOv8x-cls](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-cls.pt) | 224 | 79.0 | 94.6 | 232.0 | 1.01 | 57.4 | 154.8 |
- **acc** values are model accuracies on the [ImageNet](https://www.image-net.org/) dataset validation set. <br>Reproduce by `yolo val classify data=path/to/ImageNet device=0`
- **Speed** averaged over ImageNet val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. <br>Reproduce by `yolo val classify data=path/to/ImageNet batch=1 device=0|cpu`
</details>
## <div align="center">Integrations</div>
Our key integrations with leading AI platforms extend the functionality of Ultralytics' offerings, enhancing tasks like dataset labeling, training, visualization, and model management. Discover how Ultralytics, in collaboration with [Roboflow](https://roboflow.com/?ref=ultralytics), ClearML, [Comet](https://bit.ly/yolov8-readme-comet), Neural Magic and [OpenVINO](https://docs.ultralytics.com/integrations/openvino), can optimize your AI workflow.
<br>
<a href="https://bit.ly/ultralytics_hub" target="_blank">
<img width="100%" src="https://github.com/ultralytics/assets/raw/main/yolov8/banner-integrations.png" alt="Ultralytics active learning integrations"></a>
<br>
<br>
<div align="center">
<a href="https://roboflow.com/?ref=ultralytics">
<img src="https://github.com/ultralytics/assets/raw/main/partners/logo-roboflow.png" width="10%" alt="Roboflow logo"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="15%" height="0" alt="space">
<a href="https://clear.ml/">
<img src="https://github.com/ultralytics/assets/raw/main/partners/logo-clearml.png" width="10%" alt="ClearML logo"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="15%" height="0" alt="space">
<a href="https://bit.ly/yolov8-readme-comet">
<img src="https://github.com/ultralytics/assets/raw/main/partners/logo-comet.png" width="10%" alt="Comet ML logo"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="15%" height="0" alt="space">
<a href="https://bit.ly/yolov5-neuralmagic">
<img src="https://github.com/ultralytics/assets/raw/main/partners/logo-neuralmagic.png" width="10%" alt="NeuralMagic logo"></a>
</div>
| Roboflow | ClearML ⭐ NEW | Comet ⭐ NEW | Neural Magic ⭐ NEW |
| :--------------------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------: |
| Label and export your custom datasets directly to YOLOv8 for training with [Roboflow](https://roboflow.com/?ref=ultralytics) | Automatically track, visualize and even remotely train YOLOv8 using [ClearML](https://clear.ml/) (open-source!) | Free forever, [Comet](https://bit.ly/yolov8-readme-comet) lets you save YOLOv8 models, resume training, and interactively visualize and debug predictions | Run YOLOv8 inference up to 6x faster with [Neural Magic DeepSparse](https://bit.ly/yolov5-neuralmagic) |
## <div align="center">Ultralytics HUB</div>
Experience seamless AI with [Ultralytics HUB](https://bit.ly/ultralytics_hub) ⭐, the all-in-one solution for data visualization, YOLOv5 and YOLOv8 🚀 model training and deployment, without any coding. Transform images into actionable insights and bring your AI visions to life with ease using our cutting-edge platform and user-friendly [Ultralytics App](https://ultralytics.com/app_install). Start your journey for **Free** now!
<a href="https://bit.ly/ultralytics_hub" target="_blank">
<img width="100%" src="https://github.com/ultralytics/assets/raw/main/im/ultralytics-hub.png" alt="Ultralytics HUB preview image"></a>
## <div align="center">Contribute</div>
We love your input! YOLOv5 and YOLOv8 would not be possible without help from our community. Please see our [Contributing Guide](https://docs.ultralytics.com/help/contributing) to get started, and fill out our [Survey](https://ultralytics.com/survey?utm_source=github&utm_medium=social&utm_campaign=Survey) to send us feedback on your experience. Thank you 🙏 to all our contributors!
<!-- SVG image from https://opencollective.com/ultralytics/contributors.svg?width=990 -->
<a href="https://github.com/ultralytics/yolov5/graphs/contributors">
<img width="100%" src="https://github.com/ultralytics/assets/raw/main/im/image-contributors.png" alt="Ultralytics open-source contributors"></a>
## <div align="center">License</div>
Ultralytics offers two licensing options to accommodate diverse use cases:
- **AGPL-3.0 License**: This [OSI-approved](https://opensource.org/licenses/) open-source license is ideal for students and enthusiasts, promoting open collaboration and knowledge sharing. See the [LICENSE](https://github.com/ultralytics/ultralytics/blob/main/LICENSE) file for more details.
- **Enterprise License**: Designed for commercial use, this license permits seamless integration of Ultralytics software and AI models into commercial goods and services, bypassing the open-source requirements of AGPL-3.0. If your scenario involves embedding our solutions into a commercial offering, reach out through [Ultralytics Licensing](https://ultralytics.com/license).
## <div align="center">Contact</div>
For Ultralytics bug reports and feature requests please visit [GitHub Issues](https://github.com/ultralytics/ultralytics/issues), and join our [Discord](https://ultralytics.com/discord) community for questions and discussions!
<br>
<div align="center">
<a href="https://github.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-github.png" width="3%" alt="Ultralytics GitHub"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://www.linkedin.com/company/ultralytics/"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-linkedin.png" width="3%" alt="Ultralytics LinkedIn"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://twitter.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-twitter.png" width="3%" alt="Ultralytics Twitter"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://youtube.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-youtube.png" width="3%" alt="Ultralytics YouTube"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://www.tiktok.com/@ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-tiktok.png" width="3%" alt="Ultralytics TikTok"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://www.instagram.com/ultralytics/"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-instagram.png" width="3%" alt="Ultralytics Instagram"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://ultralytics.com/discord"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-discord.png" width="3%" alt="Ultralytics Discord"></a>
</div>

297
README.zh-CN.md
View File

@ -1,297 +0,0 @@
<div align="center">
<p>
<a href="https://www.ultralytics.com/blog/ultralytics-yolov8-turns-one-a-year-of-breakthroughs-and-innovations" target="_blank">
<img width="100%" src="https://raw.githubusercontent.com/ultralytics/assets/main/im/banner-yolo-vision-2023.png" alt="YOLO Vision banner"></a>
</p>
[中文](https://docs.ultralytics.com/zh/) | [한국어](https://docs.ultralytics.com/ko/) | [日本語](https://docs.ultralytics.com/ja/) | [Русский](https://docs.ultralytics.com/ru/) | [Deutsch](https://docs.ultralytics.com/de/) | [Français](https://docs.ultralytics.com/fr/) | [Español](https://docs.ultralytics.com/es/) | [Português](https://docs.ultralytics.com/pt/) | [हिन्दी](https://docs.ultralytics.com/hi/) | [العربية](https://docs.ultralytics.com/ar/) <br>
<div>
<a href="https://github.com/ultralytics/ultralytics/actions/workflows/ci.yaml"><img src="https://github.com/ultralytics/ultralytics/actions/workflows/ci.yaml/badge.svg" alt="Ultralytics CI"></a>
<a href="https://codecov.io/github/ultralytics/ultralytics"><img src="https://codecov.io/github/ultralytics/ultralytics/branch/main/graph/badge.svg?token=HHW7IIVFVY" alt="Ultralytics Code Coverage"></a>
<a href="https://zenodo.org/badge/latestdoi/264818686"><img src="https://zenodo.org/badge/264818686.svg" alt="YOLOv8 Citation"></a>
<a href="https://hub.docker.com/r/ultralytics/ultralytics"><img src="https://img.shields.io/docker/pulls/ultralytics/ultralytics?logo=docker" alt="Docker Pulls"></a>
<a href="https://ultralytics.com/discord"><img alt="Discord" src="https://img.shields.io/discord/1089800235347353640?logo=discord&logoColor=white&label=Discord&color=blue"></a>
<br>
<a href="https://console.paperspace.com/github/ultralytics/ultralytics"><img src="https://assets.paperspace.io/img/gradient-badge.svg" alt="Run on Gradient"></a>
<a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/tutorial.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a>
<a href="https://www.kaggle.com/ultralytics/yolov8"><img src="https://kaggle.com/static/images/open-in-kaggle.svg" alt="Open In Kaggle"></a>
</div>
<br>
[Ultralytics](https://ultralytics.com) [YOLOv8](https://github.com/ultralytics/ultralytics) 是一款前沿、最先进（SOTA）的模型，基于先前 YOLO 版本的成功，引入了新功能和改进，进一步提升性能和灵活性。YOLOv8 设计快速、准确且易于使用，使其成为各种物体检测与跟踪、实例分割、图像分类和姿态估计任务的绝佳选择。
我们希望这里的资源能帮助您充分利用 YOLOv8。请浏览 YOLOv8 <a href="https://docs.ultralytics.com/">文档</a> 了解详细信息，在 <a href="https://github.com/ultralytics/ultralytics/issues/new/choose">GitHub</a> 上提交问题以获得支持，并加入我们的 <a href="https://ultralytics.com/discord">Discord</a> 社区进行问题和讨论！
如需申请企业许可，请在 [Ultralytics Licensing](https://ultralytics.com/license) 处填写表格
<img width="100%" src="https://raw.githubusercontent.com/ultralytics/assets/main/yolov8/yolo-comparison-plots.png" alt="YOLOv8 performance plots"></a>
<div align="center">
<a href="https://github.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-github.png" width="2%" alt="Ultralytics GitHub"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://www.linkedin.com/company/ultralytics/"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-linkedin.png" width="2%" alt="Ultralytics LinkedIn"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://twitter.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-twitter.png" width="2%" alt="Ultralytics Twitter"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://youtube.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-youtube.png" width="2%" alt="Ultralytics YouTube"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://www.tiktok.com/@ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-tiktok.png" width="2%" alt="Ultralytics TikTok"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://www.instagram.com/ultralytics/"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-instagram.png" width="2%" alt="Ultralytics Instagram"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="2%" alt="space">
<a href="https://ultralytics.com/discord"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-discord.png" width="2%" alt="Ultralytics Discord"></a>
</div>
</div>
以下是提供的内容的中文翻译：
## <div align="center">文档</div>
请参阅下面的快速安装和使用示例，以及 [YOLOv8 文档](https://docs.ultralytics.com) 上有关训练、验证、预测和部署的完整文档。
<details open>
<summary>安装</summary>
使用Pip在一个[**Python>=3.8**](https://www.python.org/)环境中安装`ultralytics`包，此环境还需包含[**PyTorch>=1.8**](https://pytorch.org/get-started/locally/)。这也会安装所有必要的[依赖项](https://github.com/ultralytics/ultralytics/blob/main/pyproject.toml)。
[![PyPI version](https://badge.fury.io/py/ultralytics.svg)](https://badge.fury.io/py/ultralytics) [![Downloads](https://static.pepy.tech/badge/ultralytics)](https://pepy.tech/project/ultralytics)
```bash
pip install ultralytics
```
如需使用包括[Conda](https://anaconda.org/conda-forge/ultralytics)、[Docker](https://hub.docker.com/r/ultralytics/ultralytics)和Git在内的其他安装方法，请参考[快速入门指南](https://docs.ultralytics.com/quickstart)。
</details>
<details open>
<summary>Usage</summary>
### CLI
YOLOv8 可以在命令行界面（CLI）中直接使用，只需输入 `yolo` 命令：
```bash
yolo predict model=yolov8n.pt source='https://ultralytics.com/images/bus.jpg'
```
`yolo` 可用于各种任务和模式，并接受其他参数，例如 `imgsz=640`。查看 YOLOv8 [CLI 文档](https://docs.ultralytics.com/usage/cli)以获取示例。
### Python
YOLOv8 也可以在 Python 环境中直接使用，并接受与上述 CLI 示例中相同的[参数](https://docs.ultralytics.com/usage/cfg/)：
```python
from ultralytics import YOLO
# 加载模型
model = YOLO("yolov8n.yaml") # 从头开始构建新模型
model = YOLO("yolov8n.pt") # 加载预训练模型（建议用于训练）
# 使用模型
model.train(data="coco128.yaml", epochs=3) # 训练模型
metrics = model.val() # 在验证集上评估模型性能
results = model("https://ultralytics.com/images/bus.jpg") # 对图像进行预测
success = model.export(format="onnx") # 将模型导出为 ONNX 格式
```
查看 YOLOv8 [Python 文档](https://docs.ultralytics.com/usage/python)以获取更多示例。
</details>
### 笔记本
Ultralytics 提供了 YOLOv8 的交互式笔记本，涵盖训练、验证、跟踪等内容。每个笔记本都配有 [YouTube](https://youtube.com/ultralytics) 教程，使学习和实现高级 YOLOv8 功能变得简单。
| 文档 | 笔记本 | YouTube |
| ------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
| <a href="https://docs.ultralytics.com/modes/">YOLOv8 训练、验证、预测和导出模式</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/tutorial.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="在 Colab 中打开"></a> | <a href="https://youtu.be/j8uQc0qB91s"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube 视频"></center></a> |
| <a href="https://docs.ultralytics.com/hub/quickstart/">Ultralytics HUB 快速开始</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/hub.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="在 Colab 中打开"></a> | <a href="https://youtu.be/lveF9iCMIzc"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube 视频"></center></a> |
| <a href="https://docs.ultralytics.com/modes/track/">YOLOv8 视频中的多对象跟踪</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/object_tracking.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="在 Colab 中打开"></a> | <a href="https://youtu.be/hHyHmOtmEgs"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube 视频"></center></a> |
| <a href="https://docs.ultralytics.com/guides/object-counting/">YOLOv8 视频中的对象计数</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/object_counting.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="在 Colab 中打开"></a> | <a href="https://youtu.be/Ag2e-5_NpS0"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube 视频"></center></a> |
| <a href="https://docs.ultralytics.com/guides/heatmaps/">YOLOv8 视频中的热图</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/examples/heatmaps.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="在 Colab 中打开"></a> | <a href="https://youtu.be/4ezde5-nZZw"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube 视频"></center></a> |
| <a href="https://docs.ultralytics.com/datasets/explorer/">Ultralytics 数据集浏览器，集成 SQL 和 OpenAI 🚀 New</a> | <a href="https://colab.research.google.com/github/ultralytics/ultralytics/blob/main/docs/en/datasets/explorer/explorer.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="在 Colab 中打开"></a> | <a href="https://youtu.be/3VryynorQeo"><center><img width=30% src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-social-youtube-rect.png" alt="Ultralytics Youtube Video"></center></a> |
## <div align="center">模型</div>
在[COCO](https://docs.ultralytics.com/datasets/detect/coco)数据集上预训练的YOLOv8 [检测](https://docs.ultralytics.com/tasks/detect)，[分割](https://docs.ultralytics.com/tasks/segment)和[姿态](https://docs.ultralytics.com/tasks/pose)模型可以在这里找到，以及在[ImageNet](https://docs.ultralytics.com/datasets/classify/imagenet)数据集上预训练的YOLOv8 [分类](https://docs.ultralytics.com/tasks/classify)模型。所有的检测，分割和姿态模型都支持[追踪](https://docs.ultralytics.com/modes/track)模式。
<img width="1024" src="https://raw.githubusercontent.com/ultralytics/assets/main/im/banner-tasks.png" alt="Ultralytics YOLO supported tasks">
所有[模型](https://github.com/ultralytics/ultralytics/tree/main/ultralytics/cfg/models)在首次使用时会自动从最新的Ultralytics [发布版本](https://github.com/ultralytics/assets/releases)下载。
<details open><summary>检测 (COCO)</summary>
查看[检测文档](https://docs.ultralytics.com/tasks/detect/)以获取这些在[COCO](https://docs.ultralytics.com/datasets/detect/coco/)上训练的模型的使用示例，其中包括80个预训练类别。
| 模型 | 尺寸<br><sup>(像素) | mAP<sup>val<br>50-95 | 速度<br><sup>CPU ONNX<br>(ms) | 速度<br><sup>A100 TensorRT<br>(ms) | 参数<br><sup>(M) | FLOPs<br><sup>(B) |
| ------------------------------------------------------------------------------------ | --------------- | -------------------- | --------------------------- | -------------------------------- | -------------- | ----------------- |
| [YOLOv8n](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n.pt) | 640 | 37.3 | 80.4 | 0.99 | 3.2 | 8.7 |
| [YOLOv8s](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s.pt) | 640 | 44.9 | 128.4 | 1.20 | 11.2 | 28.6 |
| [YOLOv8m](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m.pt) | 640 | 50.2 | 234.7 | 1.83 | 25.9 | 78.9 |
| [YOLOv8l](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l.pt) | 640 | 52.9 | 375.2 | 2.39 | 43.7 | 165.2 |
| [YOLOv8x](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x.pt) | 640 | 53.9 | 479.1 | 3.53 | 68.2 | 257.8 |
- **mAP<sup>val</sup>** 值是基于单模型单尺度在 [COCO val2017](https://cocodataset.org) 数据集上的结果。 <br>通过 `yolo val detect data=coco.yaml device=0` 复现
- **速度** 是使用 [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) 实例对 COCO val 图像进行平均计算的。 <br>通过 `yolo val detect data=coco.yaml batch=1 device=0|cpu` 复现
</details>
<details><summary>检测（Open Image V7）</summary>
查看[检测文档](https://docs.ultralytics.com/tasks/detect/)以获取这些在[Open Image V7](https://docs.ultralytics.com/datasets/detect/open-images-v7/)上训练的模型的使用示例，其中包括600个预训练类别。
| 模型 | 尺寸<br><sup>(像素) | mAP<sup>验证<br>50-95 | 速度<br><sup>CPU ONNX<br>(毫秒) | 速度<br><sup>A100 TensorRT<br>(毫秒) | 参数<br><sup>(M) | 浮点运算<br><sup>(B) |
| ----------------------------------------------------------------------------------------- | --------------- | ------------------- | --------------------------- | -------------------------------- | -------------- | ---------------- |
| [YOLOv8n](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-oiv7.pt) | 640 | 18.4 | 142.4 | 1.21 | 3.5 | 10.5 |
| [YOLOv8s](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-oiv7.pt) | 640 | 27.7 | 183.1 | 1.40 | 11.4 | 29.7 |
| [YOLOv8m](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-oiv7.pt) | 640 | 33.6 | 408.5 | 2.26 | 26.2 | 80.6 |
| [YOLOv8l](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-oiv7.pt) | 640 | 34.9 | 596.9 | 2.43 | 44.1 | 167.4 |
| [YOLOv8x](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-oiv7.pt) | 640 | 36.3 | 860.6 | 3.56 | 68.7 | 260.6 |
- **mAP<sup>验证</sup>** 值适用于在[Open Image V7](https://docs.ultralytics.com/datasets/detect/open-images-v7/)数据集上的单模型单尺度。 <br>通过 `yolo val detect data=open-images-v7.yaml device=0` 以复现。
- **速度** 在使用[Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/)实例对Open Image V7验证图像进行平均测算。 <br>通过 `yolo val detect data=open-images-v7.yaml batch=1 device=0|cpu` 以复现。
</details>
<details><summary>分割 (COCO)</summary>
查看[分割文档](https://docs.ultralytics.com/tasks/segment/)以获取这些在[COCO-Seg](https://docs.ultralytics.com/datasets/segment/coco/)上训练的模型的使用示例，其中包括80个预训练类别。
| 模型 | 尺寸<br><sup>(像素) | mAP<sup>box<br>50-95 | mAP<sup>mask<br>50-95 | 速度<br><sup>CPU ONNX<br>(ms) | 速度<br><sup>A100 TensorRT<br>(ms) | 参数<br><sup>(M) | FLOPs<br><sup>(B) |
| -------------------------------------------------------------------------------------------- | --------------- | -------------------- | --------------------- | --------------------------- | -------------------------------- | -------------- | ----------------- |
| [YOLOv8n-seg](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-seg.pt) | 640 | 36.7 | 30.5 | 96.1 | 1.21 | 3.4 | 12.6 |
| [YOLOv8s-seg](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-seg.pt) | 640 | 44.6 | 36.8 | 155.7 | 1.47 | 11.8 | 42.6 |
| [YOLOv8m-seg](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-seg.pt) | 640 | 49.9 | 40.8 | 317.0 | 2.18 | 27.3 | 110.2 |
| [YOLOv8l-seg](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-seg.pt) | 640 | 52.3 | 42.6 | 572.4 | 2.79 | 46.0 | 220.5 |
| [YOLOv8x-seg](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-seg.pt) | 640 | 53.4 | 43.4 | 712.1 | 4.02 | 71.8 | 344.1 |
- **mAP<sup>val</sup>** 值是基于单模型单尺度在 [COCO val2017](https://cocodataset.org) 数据集上的结果。 <br>通过 `yolo val segment data=coco-seg.yaml device=0` 复现
- **速度** 是使用 [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) 实例对 COCO val 图像进行平均计算的。 <br>通过 `yolo val segment data=coco-seg.yaml batch=1 device=0|cpu` 复现
</details>
<details><summary>姿态 (COCO)</summary>
查看[姿态文档](https://docs.ultralytics.com/tasks/pose/)以获取这些在[COCO-Pose](https://docs.ultralytics.com/datasets/pose/coco/)上训练的模型的使用示例，其中包括1个预训练类别，即人。
| 模型 | 尺寸<br><sup>(像素) | mAP<sup>pose<br>50-95 | mAP<sup>pose<br>50 | 速度<br><sup>CPU ONNX<br>(ms) | 速度<br><sup>A100 TensorRT<br>(ms) | 参数<br><sup>(M) | FLOPs<br><sup>(B) |
| ---------------------------------------------------------------------------------------------------- | --------------- | --------------------- | ------------------ | --------------------------- | -------------------------------- | -------------- | ----------------- |
| [YOLOv8n-pose](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-pose.pt) | 640 | 50.4 | 80.1 | 131.8 | 1.18 | 3.3 | 9.2 |
| [YOLOv8s-pose](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-pose.pt) | 640 | 60.0 | 86.2 | 233.2 | 1.42 | 11.6 | 30.2 |
| [YOLOv8m-pose](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-pose.pt) | 640 | 65.0 | 88.8 | 456.3 | 2.00 | 26.4 | 81.0 |
| [YOLOv8l-pose](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-pose.pt) | 640 | 67.6 | 90.0 | 784.5 | 2.59 | 44.4 | 168.6 |
| [YOLOv8x-pose](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-pose.pt) | 640 | 69.2 | 90.2 | 1607.1 | 3.73 | 69.4 | 263.2 |
| [YOLOv8x-pose-p6](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-pose-p6.pt) | 1280 | 71.6 | 91.2 | 4088.7 | 10.04 | 99.1 | 1066.4 |
- **mAP<sup>val</sup>** 值是基于单模型单尺度在 [COCO Keypoints val2017](https://cocodataset.org) 数据集上的结果。 <br>通过 `yolo val pose data=coco-pose.yaml device=0` 复现
- **速度** 是使用 [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) 实例对 COCO val 图像进行平均计算的。 <br>通过 `yolo val pose data=coco-pose.yaml batch=1 device=0|cpu` 复现
</details>
<details><summary>旋转检测 (DOTAv1)</summary>
查看[旋转检测文档](https://docs.ultralytics.com/tasks/obb/)以获取这些在[DOTAv1](https://docs.ultralytics.com/datasets/obb/dota-v2/#dota-v10/)上训练的模型的使用示例，其中包括15个预训练类别。
| 模型 | 尺寸<br><sup>(像素) | mAP<sup>test<br>50 | 速度<br><sup>CPU ONNX<br>(ms) | 速度<br><sup>A100 TensorRT<br>(ms) | 参数<br><sup>(M) | FLOPs<br><sup>(B) |
| -------------------------------------------------------------------------------------------- | --------------- | ------------------ | --------------------------- | -------------------------------- | -------------- | ----------------- |
| [YOLOv8n-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-obb.pt) | 1024 | 78.0 | 204.77 | 3.57 | 3.1 | 23.3 |
| [YOLOv8s-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-obb.pt) | 1024 | 79.5 | 424.88 | 4.07 | 11.4 | 76.3 |
| [YOLOv8m-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-obb.pt) | 1024 | 80.5 | 763.48 | 7.61 | 26.4 | 208.6 |
| [YOLOv8l-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-obb.pt) | 1024 | 80.7 | 1278.42 | 11.83 | 44.5 | 433.8 |
| [YOLOv8x-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-obb.pt) | 1024 | 81.36 | 1759.10 | 13.23 | 69.5 | 676.7 |
- **mAP<sup>val</sup>** 值是基于单模型多尺度在 [DOTAv1](https://captain-whu.github.io/DOTA/index.html) 数据集上的结果。 <br>通过 `yolo val obb data=DOTAv1.yaml device=0 split=test` 复现
- **速度** 是使用 [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) 实例对 COCO val 图像进行平均计算的。 <br>通过 `yolo val obb data=DOTAv1.yaml batch=1 device=0|cpu` 复现
</details>
<details><summary>分类 (ImageNet)</summary>
查看[分类文档](https://docs.ultralytics.com/tasks/classify/)以获取这些在[ImageNet](https://docs.ultralytics.com/datasets/classify/imagenet/)上训练的模型的使用示例，其中包括1000个预训练类别。
| 模型 | 尺寸<br><sup>(像素) | acc<br><sup>top1 | acc<br><sup>top5 | 速度<br><sup>CPU ONNX<br>(ms) | 速度<br><sup>A100 TensorRT<br>(ms) | 参数<br><sup>(M) | FLOPs<br><sup>(B) at 640 |
| -------------------------------------------------------------------------------------------- | --------------- | ---------------- | ---------------- | --------------------------- | -------------------------------- | -------------- | ------------------------ |
| [YOLOv8n-cls](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-cls.pt) | 224 | 69.0 | 88.3 | 12.9 | 0.31 | 2.7 | 4.3 |
| [YOLOv8s-cls](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-cls.pt) | 224 | 73.8 | 91.7 | 23.4 | 0.35 | 6.4 | 13.5 |
| [YOLOv8m-cls](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-cls.pt) | 224 | 76.8 | 93.5 | 85.4 | 0.62 | 17.0 | 42.7 |
| [YOLOv8l-cls](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-cls.pt) | 224 | 76.8 | 93.5 | 163.0 | 0.87 | 37.5 | 99.7 |
| [YOLOv8x-cls](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-cls.pt) | 224 | 79.0 | 94.6 | 232.0 | 1.01 | 57.4 | 154.8 |
- **acc** 值是模型在 [ImageNet](https://www.image-net.org/) 数据集验证集上的准确率。 <br>通过 `yolo val classify data=path/to/ImageNet device=0` 复现
- **速度** 是使用 [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) 实例对 ImageNet val 图像进行平均计算的。 <br>通过 `yolo val classify data=path/to/ImageNet batch=1 device=0|cpu` 复现
</details>
## <div align="center">集成</div>
我们与领先的AI平台的关键整合扩展了Ultralytics产品的功能，增强了数据集标签化、训练、可视化和模型管理等任务。探索Ultralytics如何与[Roboflow](https://roboflow.com/?ref=ultralytics)、ClearML、[Comet](https://bit.ly/yolov8-readme-comet)、Neural Magic以及[OpenVINO](https://docs.ultralytics.com/integrations/openvino)合作，优化您的AI工作流程。
<br>
<a href="https://bit.ly/ultralytics_hub" target="_blank">
<img width="100%" src="https://github.com/ultralytics/assets/raw/main/yolov8/banner-integrations.png" alt="Ultralytics active learning integrations"></a>
<br>
<br>
<div align="center">
<a href="https://roboflow.com/?ref=ultralytics">
<img src="https://github.com/ultralytics/assets/raw/main/partners/logo-roboflow.png" width="10%" alt="Roboflow logo"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="15%" height="0" alt="space">
<a href="https://clear.ml/">
<img src="https://github.com/ultralytics/assets/raw/main/partners/logo-clearml.png" width="10%" alt="ClearML logo"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="15%" height="0" alt="space">
<a href="https://bit.ly/yolov8-readme-comet">
<img src="https://github.com/ultralytics/assets/raw/main/partners/logo-comet.png" width="10%" alt="Comet ML logo"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="15%" height="0" alt="space">
<a href="https://bit.ly/yolov5-neuralmagic">
<img src="https://github.com/ultralytics/assets/raw/main/partners/logo-neuralmagic.png" width="10%" alt="NeuralMagic logo"></a>
</div>
| Roboflow | ClearML ⭐ NEW | Comet ⭐ NEW | Neural Magic ⭐ NEW |
| :--------------------------------------------------------------------------------: | :--------------------------------------------------------: | :----------------------------------------------------------------------------------: | :-----------------------------------------------------------------------------------: |
| 使用 [Roboflow](https://roboflow.com/?ref=ultralytics) 将您的自定义数据集直接标记并导出至 YOLOv8 进行训练 | 使用 [ClearML](https://clear.ml/)（开源！）自动跟踪、可视化，甚至远程训练 YOLOv8 | 免费且永久，[Comet](https://bit.ly/yolov8-readme-comet) 让您保存 YOLOv8 模型、恢复训练，并以交互式方式查看和调试预测 | 使用 [Neural Magic DeepSparse](https://bit.ly/yolov5-neuralmagic) 使 YOLOv8 推理速度提高多达 6 倍 |
## <div align="center">Ultralytics HUB</div>
体验 [Ultralytics HUB](https://bit.ly/ultralytics_hub) ⭐ 带来的无缝 AI，这是一个一体化解决方案，用于数据可视化、YOLOv5 和即将推出的 YOLOv8 🚀 模型训练和部署，无需任何编码。通过我们先进的平台和用户友好的 [Ultralytics 应用程序](https://ultralytics.com/app_install)，轻松将图像转化为可操作的见解，并实现您的 AI 愿景。现在就开始您的**免费**之旅！
<a href="https://bit.ly/ultralytics_hub" target="_blank">
<img width="100%" src="https://github.com/ultralytics/assets/raw/main/im/ultralytics-hub.png" alt="Ultralytics HUB preview image"></a>
## <div align="center">贡献</div>
我们喜欢您的参与！没有社区的帮助，YOLOv5 和 YOLOv8 将无法实现。请参阅我们的[贡献指南](https://docs.ultralytics.com/help/contributing)以开始使用，并填写我们的[调查问卷](https://ultralytics.com/survey?utm_source=github&utm_medium=social&utm_campaign=Survey)向我们提供您的使用体验反馈。感谢所有贡献者的支持！🙏
<!-- SVG image from https://opencollective.com/ultralytics/contributors.svg?width=990 -->
<a href="https://github.com/ultralytics/yolov5/graphs/contributors">
<img width="100%" src="https://github.com/ultralytics/assets/raw/main/im/image-contributors.png" alt="Ultralytics open-source contributors"></a>
## <div align="center">许可证</div>
Ultralytics 提供两种许可证选项以适应各种使用场景：
- **AGPL-3.0 许可证**：这个[OSI 批准](https://opensource.org/licenses/)的开源许可证非常适合学生和爱好者，可以推动开放的协作和知识分享。请查看[LICENSE](https://github.com/ultralytics/ultralytics/blob/main/LICENSE) 文件以了解更多细节。
- **企业许可证**：专为商业用途设计，该许可证允许将 Ultralytics 的软件和 AI 模型无缝集成到商业产品和服务中，从而绕过 AGPL-3.0 的开源要求。如果您的场景涉及将我们的解决方案嵌入到商业产品中，请通过 [Ultralytics Licensing](https://ultralytics.com/license)与我们联系。
## <div align="center">联系方式</div>
对于 Ultralytics 的错误报告和功能请求，请访问 [GitHub Issues](https://github.com/ultralytics/ultralytics/issues)，并加入我们的 [Discord](https://ultralytics.com/discord) 社区进行问题和讨论！
<br>
<div align="center">
<a href="https://github.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-github.png" width="3%" alt="Ultralytics GitHub"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://www.linkedin.com/company/ultralytics/"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-linkedin.png" width="3%" alt="Ultralytics LinkedIn"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://twitter.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-twitter.png" width="3%" alt="Ultralytics Twitter"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://youtube.com/ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-youtube.png" width="3%" alt="Ultralytics YouTube"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://www.tiktok.com/@ultralytics"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-tiktok.png" width="3%" alt="Ultralytics TikTok"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://www.instagram.com/ultralytics/"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-instagram.png" width="3%" alt="Ultralytics Instagram"></a>
<img src="https://github.com/ultralytics/assets/raw/main/social/logo-transparent.png" width="3%" alt="space">
<a href="https://ultralytics.com/discord"><img src="https://github.com/ultralytics/assets/raw/main/social/logo-social-discord.png" width="3%" alt="Ultralytics Discord"></a>
</div>

2140
figures/latency.svg Normal file
View File

File diff suppressed because it is too large Load Diff
Side by Side

After

Width:  |  Height:  |  Size: 59 KiB

2108
figures/params.svg Normal file
View File

File diff suppressed because it is too large Load Diff
Side by Side

After

Width:  |  Height:  |  Size: 58 KiB

9
requirements.txt Normal file
View File

@ -0,0 +1,9 @@
torch
torchvision
onnx
onnxruntime
pycocotools
PyYAML
scipy
onnxsim
onnxruntime-gpu

3
ultralytics/__init__.py
View File

@ -3,7 +3,7 @@
__version__ = "8.1.34" __version__ = "8.1.34"
from ultralytics.data.explorer.explorer import Explorer from ultralytics.data.explorer.explorer import Explorer
from ultralytics.models import RTDETR, SAM, YOLO, YOLOWorld from ultralytics.models import RTDETR, SAM, YOLO, YOLOWorld, YOLOv10
from ultralytics.models.fastsam import FastSAM from ultralytics.models.fastsam import FastSAM
from ultralytics.models.nas import NAS from ultralytics.models.nas import NAS
from ultralytics.utils import ASSETS, SETTINGS as settings from ultralytics.utils import ASSETS, SETTINGS as settings
@ -23,4 +23,5 @@ __all__ = (
"download", "download",
"settings", "settings",
"Explorer", "Explorer",
"YOLOv10"
) )

4
ultralytics/cfg/__init__.py
View File

@ -549,6 +549,10 @@ def entrypoint(debug=""):
from ultralytics import SAM from ultralytics import SAM
model = SAM(model) model = SAM(model)
elif "yolov10" in stem:
from ultralytics import YOLOv10
model = YOLOv10(model)
else: else:
from ultralytics import YOLO from ultralytics import YOLO

40
ultralytics/cfg/models/v10/yolov10b.yaml Normal file
View File

@ -0,0 +1,40 @@
# Parameters
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
b: [0.67, 1.00, 512]
# YOLOv8.0n backbone
backbone:
# [from, repeats, module, args]
- [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, SCDown, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, SCDown, [1024, 3, 2]] # 7-P5/32
- [-1, 3, C2fCIB, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 9
- [-1, 1, PSA, [1024]] # 10
# YOLOv8.0n head
head:
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 6], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2fCIB, [512, True]] # 13
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 4], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 16 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 13], 1, Concat, [1]] # cat head P4
- [-1, 3, C2fCIB, [512, True]] # 19 (P4/16-medium)
- [-1, 1, SCDown, [512, 3, 2]]
- [[-1, 10], 1, Concat, [1]] # cat head P5
- [-1, 3, C2fCIB, [1024, True]] # 22 (P5/32-large)
- [[16, 19, 22], 1, v10Detect, [nc]] # Detect(P3, P4, P5)

40
ultralytics/cfg/models/v10/yolov10l.yaml Normal file
View File

@ -0,0 +1,40 @@
# Parameters
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
l: [1.00, 1.00, 512] # YOLOv8l summary: 365 layers, 43691520 parameters, 43691504 gradients, 165.7 GFLOPs
# YOLOv8.0n backbone
backbone:
# [from, repeats, module, args]
- [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, SCDown, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, SCDown, [1024, 3, 2]] # 7-P5/32
- [-1, 3, C2fCIB, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 9
- [-1, 1, PSA, [1024]] # 10
# YOLOv8.0n head
head:
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 6], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2fCIB, [512, True]] # 13
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 4], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 16 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 13], 1, Concat, [1]] # cat head P4
- [-1, 3, C2fCIB, [512, True]] # 19 (P4/16-medium)
- [-1, 1, SCDown, [512, 3, 2]]
- [[-1, 10], 1, Concat, [1]] # cat head P5
- [-1, 3, C2fCIB, [1024, True]] # 22 (P5/32-large)
- [[16, 19, 22], 1, v10Detect, [nc]] # Detect(P3, P4, P5)

43
ultralytics/cfg/models/v10/yolov10m.yaml Normal file
View File

@ -0,0 +1,43 @@
# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLOv8 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect
# Parameters
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
m: [0.67, 0.75, 768] # YOLOv8m summary: 295 layers, 25902640 parameters, 25902624 gradients, 79.3 GFLOPs
# YOLOv8.0n backbone
backbone:
# [from, repeats, module, args]
- [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, SCDown, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, SCDown, [1024, 3, 2]] # 7-P5/32
- [-1, 3, C2fCIB, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 9
- [-1, 1, PSA, [1024]] # 10
# YOLOv8.0n head
head:
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 6], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [512]] # 13
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 4], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 16 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 13], 1, Concat, [1]] # cat head P4
- [-1, 3, C2fCIB, [512, True]] # 19 (P4/16-medium)
- [-1, 1, SCDown, [512, 3, 2]]
- [[-1, 10], 1, Concat, [1]] # cat head P5
- [-1, 3, C2fCIB, [1024, True]] # 22 (P5/32-large)
- [[16, 19, 22], 1, v10Detect, [nc]] # Detect(P3, P4, P5)

40
ultralytics/cfg/models/v10/yolov10n.yaml Normal file
View File

@ -0,0 +1,40 @@
# Parameters
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
n: [0.33, 0.25, 1024]
# YOLOv8.0n backbone
backbone:
# [from, repeats, module, args]
- [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, SCDown, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, SCDown, [1024, 3, 2]] # 7-P5/32
- [-1, 3, C2f, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 9
- [-1, 1, PSA, [1024]] # 10
# YOLOv8.0n head
head:
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 6], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [512]] # 13
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 4], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 16 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 13], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [512]] # 19 (P4/16-medium)
- [-1, 1, SCDown, [512, 3, 2]]
- [[-1, 10], 1, Concat, [1]] # cat head P5
- [-1, 3, C2fCIB, [1024, True, True]] # 22 (P5/32-large)
- [[16, 19, 22], 1, v10Detect, [nc]] # Detect(P3, P4, P5)

39
ultralytics/cfg/models/v10/yolov10s.yaml Normal file
View File

@ -0,0 +1,39 @@
# Parameters
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
s: [0.33, 0.50, 1024]
backbone:
# [from, repeats, module, args]
- [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, SCDown, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, SCDown, [1024, 3, 2]] # 7-P5/32
- [-1, 3, C2fCIB, [1024, True, True]]
- [-1, 1, SPPF, [1024, 5]] # 9
- [-1, 1, PSA, [1024]] # 10
# YOLOv8.0n head
head:
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 6], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [512]] # 13
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 4], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 16 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 13], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [512]] # 19 (P4/16-medium)
- [-1, 1, SCDown, [512, 3, 2]]
- [[-1, 10], 1, Concat, [1]] # cat head P5
- [-1, 3, C2fCIB, [1024, True, True]] # 22 (P5/32-large)
- [[16, 19, 22], 1, v10Detect, [nc]] # Detect(P3, P4, P5)

40
ultralytics/cfg/models/v10/yolov10x.yaml Normal file
View File

@ -0,0 +1,40 @@
# Parameters
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
x: [1.00, 1.25, 512]
# YOLOv8.0n backbone
backbone:
# [from, repeats, module, args]
- [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, SCDown, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2fCIB, [512, True]]
- [-1, 1, SCDown, [1024, 3, 2]] # 7-P5/32
- [-1, 3, C2fCIB, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 9
- [-1, 1, PSA, [1024]] # 10
# YOLOv8.0n head
head:
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 6], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2fCIB, [512, True]] # 13
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 4], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 16 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 13], 1, Concat, [1]] # cat head P4
- [-1, 3, C2fCIB, [512, True]] # 19 (P4/16-medium)
- [-1, 1, SCDown, [512, 3, 2]]
- [[-1, 10], 1, Concat, [1]] # cat head P5
- [-1, 3, C2fCIB, [1024, True]] # 22 (P5/32-large)
- [[16, 19, 22], 1, v10Detect, [nc]] # Detect(P3, P4, P5)

5
ultralytics/engine/exporter.py
View File

@ -67,7 +67,7 @@ from ultralytics.cfg import get_cfg
from ultralytics.data.dataset import YOLODataset from ultralytics.data.dataset import YOLODataset
from ultralytics.data.utils import check_det_dataset from ultralytics.data.utils import check_det_dataset
from ultralytics.nn.autobackend import check_class_names, default_class_names from ultralytics.nn.autobackend import check_class_names, default_class_names
from ultralytics.nn.modules import C2f, Detect, RTDETRDecoder from ultralytics.nn.modules import C2f, Detect, RTDETRDecoder, v10Detect
from ultralytics.nn.tasks import DetectionModel, SegmentationModel, WorldModel from ultralytics.nn.tasks import DetectionModel, SegmentationModel, WorldModel
from ultralytics.utils import ( from ultralytics.utils import (
ARM64, ARM64,
@ -229,6 +229,9 @@ class Exporter:
m.dynamic = self.args.dynamic m.dynamic = self.args.dynamic
m.export = True m.export = True
m.format = self.args.format m.format = self.args.format
if isinstance(m, v10Detect):
m.max_det = self.args.max_det
elif isinstance(m, C2f) and not any((saved_model, pb, tflite, edgetpu, tfjs)): elif isinstance(m, C2f) and not any((saved_model, pb, tflite, edgetpu, tfjs)):
# EdgeTPU does not support FlexSplitV while split provides cleaner ONNX graph # EdgeTPU does not support FlexSplitV while split provides cleaner ONNX graph
m.forward = m.forward_split m.forward = m.forward_split

3
ultralytics/engine/trainer.py
View File

@ -425,7 +425,8 @@ class BaseTrainer:
self.ema.update_attr(self.model, include=["yaml", "nc", "args", "names", "stride", "class_weights"]) self.ema.update_attr(self.model, include=["yaml", "nc", "args", "names", "stride", "class_weights"])
# Validation # Validation
if self.args.val or final_epoch or self.stopper.possible_stop or self.stop: if (self.args.val and (((epoch+1) % 10 == 0) or (self.epochs - epoch) <= 10)) \
or final_epoch or self.stopper.possible_stop or self.stop:
self.metrics, self.fitness = self.validate() self.metrics, self.fitness = self.validate()
self.save_metrics(metrics={**self.label_loss_items(self.tloss), **self.metrics, **self.lr}) self.save_metrics(metrics={**self.label_loss_items(self.tloss), **self.metrics, **self.lr})
self.stop |= self.stopper(epoch + 1, self.fitness) or final_epoch self.stop |= self.stopper(epoch + 1, self.fitness) or final_epoch

8
ultralytics/engine/validator.py
View File

@ -196,10 +196,16 @@ class BaseValidator:
self.check_stats(stats) self.check_stats(stats)
self.speed = dict(zip(self.speed.keys(), (x.t / len(self.dataloader.dataset) * 1e3 for x in dt))) self.speed = dict(zip(self.speed.keys(), (x.t / len(self.dataloader.dataset) * 1e3 for x in dt)))
self.finalize_metrics() self.finalize_metrics()
self.print_results() # self.print_results()
self.run_callbacks("on_val_end") self.run_callbacks("on_val_end")
if self.training: if self.training:
model.float() model.float()
assert(self.args.save_json and self.jdict)
with open(str(self.save_dir / "predictions.json"), "w") as f:
LOGGER.info(f"Saving {f.name}...")
json.dump(self.jdict, f) # flatten and save
stats = self.eval_json(stats) # update stats
stats['fitness'] = stats['metrics/mAP50-95(B)']
results = {**stats, **trainer.label_loss_items(self.loss.cpu() / len(self.dataloader), prefix="val")} results = {**stats, **trainer.label_loss_items(self.loss.cpu() / len(self.dataloader), prefix="val")}
return {k: round(float(v), 5) for k, v in results.items()} # return results as 5 decimal place floats return {k: round(float(v), 5) for k, v in results.items()} # return results as 5 decimal place floats
else: else:

3
ultralytics/models/__init__.py
View File

@ -3,5 +3,6 @@
from .rtdetr import RTDETR from .rtdetr import RTDETR
from .sam import SAM from .sam import SAM
from .yolo import YOLO, YOLOWorld from .yolo import YOLO, YOLOWorld
from .yolov10 import YOLOv10
__all__ = "YOLO", "RTDETR", "SAM", "YOLOWorld" # allow simpler import __all__ = "YOLO", "RTDETR", "SAM", "YOLOWorld", "YOLOv10" # allow simpler import

2
ultralytics/models/yolo/detect/val.py
View File

@ -67,7 +67,7 @@ class DetectionValidator(BaseValidator):
val = self.data.get(self.args.split, "") # validation path val = self.data.get(self.args.split, "") # validation path
self.is_coco = isinstance(val, str) and "coco" in val and val.endswith(f"{os.sep}val2017.txt") # is COCO self.is_coco = isinstance(val, str) and "coco" in val and val.endswith(f"{os.sep}val2017.txt") # is COCO
self.class_map = converter.coco80_to_coco91_class() if self.is_coco else list(range(1000)) self.class_map = converter.coco80_to_coco91_class() if self.is_coco else list(range(1000))
self.args.save_json |= self.is_coco and not self.training # run on final val if training COCO self.args.save_json |= self.is_coco # run on final val if training COCO
self.names = model.names self.names = model.names
self.nc = len(model.names) self.nc = len(model.names)
self.metrics.names = self.names self.metrics.names = self.names

5
ultralytics/models/yolov10/__init__.py Normal file
View File

@ -0,0 +1,5 @@
from .model import YOLOv10
from .predict import YOLOv10DetectionPredictor
from .val import YOLOv10DetectionValidator
__all__ = "YOLOv10DetectionPredictor", "YOLOv10DetectionValidator", "YOLOv10"

18
ultralytics/models/yolov10/model.py Normal file
View File

@ -0,0 +1,18 @@
from ..yolo import YOLO
from ultralytics.nn.tasks import YOLOv10DetectionModel
from .val import YOLOv10DetectionValidator
from .predict import YOLOv10DetectionPredictor
from .train import YOLOv10DetectionTrainer
class YOLOv10(YOLO):
@property
def task_map(self):
"""Map head to model, trainer, validator, and predictor classes."""
return {
"detect": {
"model": YOLOv10DetectionModel,
"trainer": YOLOv10DetectionTrainer,
"validator": YOLOv10DetectionValidator,
"predictor": YOLOv10DetectionPredictor,
},
}

38
ultralytics/models/yolov10/predict.py Normal file
View File

@ -0,0 +1,38 @@
from ultralytics.models.yolo.detect import DetectionPredictor
import torch
from ultralytics.utils import ops
from ultralytics.engine.results import Results
class YOLOv10DetectionPredictor(DetectionPredictor):
def postprocess(self, preds, img, orig_imgs):
if isinstance(preds, dict):
preds = preds["one2one"]
if isinstance(preds, (list, tuple)):
preds = preds[0]
if preds.shape[-1] == 6:
pass
else:
preds = preds.transpose(-1, -2)
bboxes, scores, labels = ops.v10postprocess(preds, self.args.max_det)
bboxes = ops.xywh2xyxy(bboxes)
preds = torch.cat([bboxes, scores.unsqueeze(-1), labels.unsqueeze(-1)], dim=-1)
mask = preds[..., 4] > self.args.conf
b, _, c = preds.shape
preds = preds.view(-1, preds.shape[-1])[mask.view(-1)]
preds = preds.view(b, -1, c)
if not isinstance(orig_imgs, list): # input images are a torch.Tensor, not a list
orig_imgs = ops.convert_torch2numpy_batch(orig_imgs)
results = []
for i, pred in enumerate(preds):
orig_img = orig_imgs[i]
pred[:, :4] = ops.scale_boxes(img.shape[2:], pred[:, :4], orig_img.shape)
img_path = self.batch[0][i]
results.append(Results(orig_img, path=img_path, names=self.model.names, boxes=pred))
return results

20
ultralytics/models/yolov10/train.py Normal file
View File

@ -0,0 +1,20 @@
from ultralytics.models.yolo.detect import DetectionTrainer
from .val import YOLOv10DetectionValidator
from .model import YOLOv10DetectionModel
from copy import copy
from ultralytics.utils import RANK
class YOLOv10DetectionTrainer(DetectionTrainer):
def get_validator(self):
"""Returns a DetectionValidator for YOLO model validation."""
self.loss_names = "box_om", "cls_om", "dfl_om", "box_oo", "cls_oo", "dfl_oo",
return YOLOv10DetectionValidator(
self.test_loader, save_dir=self.save_dir, args=copy(self.args), _callbacks=self.callbacks
)
def get_model(self, cfg=None, weights=None, verbose=True):
"""Return a YOLO detection model."""
model = YOLOv10DetectionModel(cfg, nc=self.data["nc"], verbose=verbose and RANK == -1)
if weights:
model.load(weights)
return model

20
ultralytics/models/yolov10/val.py Normal file
View File

@ -0,0 +1,20 @@
from ultralytics.models.yolo.detect import DetectionValidator
from ultralytics.utils import ops
import torch
class YOLOv10DetectionValidator(DetectionValidator):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.args.save_json |= self.is_coco
def postprocess(self, preds):
if isinstance(preds, dict):
preds = preds["one2one"]
if isinstance(preds, (list, tuple)):
preds = preds[0]
preds = preds.transpose(-1, -2)
boxes, scores, labels = ops.v10postprocess(preds, self.args.max_det)
bboxes = ops.xywh2xyxy(boxes)
return torch.cat([bboxes, scores.unsqueeze(-1), labels.unsqueeze(-1)], dim=-1)

11
ultralytics/nn/modules/__init__.py
View File

@ -46,6 +46,10 @@ from .block import (
CBFuse, CBFuse,
CBLinear, CBLinear,
Silence, Silence,
PSA,
C2fCIB,
SCDown,
RepVGGDW
) )
from .conv import ( from .conv import (
CBAM, CBAM,
@ -62,7 +66,7 @@ from .conv import (
RepConv, RepConv,
SpatialAttention, SpatialAttention,
) )
from .head import OBB, Classify, Detect, Pose, RTDETRDecoder, Segment, WorldDetect from .head import OBB, Classify, Detect, Pose, RTDETRDecoder, Segment, WorldDetect, v10Detect
from .transformer import ( from .transformer import (
AIFI, AIFI,
MLP, MLP,
@ -135,4 +139,9 @@ __all__ = (
"CBFuse", "CBFuse",
"CBLinear", "CBLinear",
"Silence", "Silence",
"PSA",
"C2fCIB",
"SCDown",
"RepVGGDW",
"v10Detect"
) )

129
ultralytics/nn/modules/block.py
View File

@ -7,6 +7,7 @@ import torch.nn.functional as F
from .conv import Conv, DWConv, GhostConv, LightConv, RepConv, autopad from .conv import Conv, DWConv, GhostConv, LightConv, RepConv, autopad
from .transformer import TransformerBlock from .transformer import TransformerBlock
from ultralytics.utils.torch_utils import fuse_conv_and_bn
__all__ = ( __all__ = (
"DFL", "DFL",
@ -696,3 +697,131 @@ class CBFuse(nn.Module):
res = [F.interpolate(x[self.idx[i]], size=target_size, mode="nearest") for i, x in enumerate(xs[:-1])] res = [F.interpolate(x[self.idx[i]], size=target_size, mode="nearest") for i, x in enumerate(xs[:-1])]
out = torch.sum(torch.stack(res + xs[-1:]), dim=0) out = torch.sum(torch.stack(res + xs[-1:]), dim=0)
return out return out
class RepVGGDW(torch.nn.Module):
def __init__(self, ed) -> None:
super().__init__()
self.conv = Conv(ed, ed, 7, 1, 3, g=ed, act=False)
self.conv1 = Conv(ed, ed, 3, 1, 1, g=ed, act=False)
self.dim = ed
self.act = nn.SiLU()
def forward(self, x):
return self.act(self.conv(x) + self.conv1(x))
def forward_fuse(self, x):
return self.act(self.conv(x))
@torch.no_grad()
def fuse(self):
conv = fuse_conv_and_bn(self.conv.conv, self.conv.bn)
conv1 = fuse_conv_and_bn(self.conv1.conv, self.conv1.bn)
conv_w = conv.weight
conv_b = conv.bias
conv1_w = conv1.weight
conv1_b = conv1.bias
conv1_w = torch.nn.functional.pad(conv1_w, [2,2,2,2])
final_conv_w = conv_w + conv1_w
final_conv_b = conv_b + conv1_b
conv.weight.data.copy_(final_conv_w)
conv.bias.data.copy_(final_conv_b)
self.conv = conv
del self.conv1
class CIB(nn.Module):
"""Standard bottleneck."""
def __init__(self, c1, c2, shortcut=True, e=0.5, lk=False):
"""Initializes a bottleneck module with given input/output channels, shortcut option, group, kernels, and
expansion.
"""
super().__init__()
c_ = int(c2 * e) # hidden channels
self.cv1 = nn.Sequential(
Conv(c1, c1, 3, g=c1),
Conv(c1, 2 * c_, 1),
Conv(2 * c_, 2 * c_, 3, g=2 * c_) if not lk else RepVGGDW(2 * c_),
Conv(2 * c_, c2, 1),
Conv(c2, c2, 3, g=c2),
)
self.add = shortcut and c1 == c2
def forward(self, x):
"""'forward()' applies the YOLO FPN to input data."""
return x + self.cv1(x) if self.add else self.cv1(x)
class C2fCIB(C2f):
"""Faster Implementation of CSP Bottleneck with 2 convolutions."""
def __init__(self, c1, c2, n=1, shortcut=False, lk=False, g=1, e=0.5):
"""Initialize CSP bottleneck layer with two convolutions with arguments ch_in, ch_out, number, shortcut, groups,
expansion.
"""
super().__init__(c1, c2, n, shortcut, g, e)
self.m = nn.ModuleList(CIB(self.c, self.c, shortcut, e=1.0, lk=lk) for _ in range(n))
class Attention(nn.Module):
def __init__(self, dim, num_heads=8,
attn_ratio=0.5):
super().__init__()
self.num_heads = num_heads
self.head_dim = dim // num_heads
self.key_dim = int(self.head_dim * attn_ratio)
self.scale = self.key_dim ** -0.5
nh_kd = nh_kd = self.key_dim * num_heads
h = dim + nh_kd * 2
self.qkv = Conv(dim, h, 1, act=False)
self.proj = Conv(dim, dim, 1, act=False)
self.pe = Conv(dim, dim, 3, 1, g=dim, act=False)
def forward(self, x):
B, _, H, W = x.shape
N = H * W
qkv = self.qkv(x)
q, k, v = qkv.view(B, self.num_heads, -1, N).split([self.key_dim, self.key_dim, self.head_dim], dim=2)
attn = (
(q.transpose(-2, -1) @ k) * self.scale
)
attn = attn.softmax(dim=-1)
x = (v @ attn.transpose(-2, -1)).view(B, -1, H, W) + self.pe(v.reshape(B, -1, H, W))
x = self.proj(x)
return x
class PSA(nn.Module):
def __init__(self, c1, c2, e=0.5):
super().__init__()
assert(c1 == c2)
self.c = int(c1 * e)
self.cv1 = Conv(c1, 2 * self.c, 1, 1)
self.cv2 = Conv(2 * self.c, c1, 1)
self.attn = Attention(self.c, attn_ratio=0.5, num_heads=self.c // 64)
self.ffn = nn.Sequential(
Conv(self.c, self.c*2, 1),
Conv(self.c*2, self.c, 1, act=False)
)
def forward(self, x):
a, b = self.cv1(x).split((self.c, self.c), dim=1)
b = b + self.attn(b)
b = b + self.ffn(b)
return self.cv2(torch.cat((a, b), 1))
class SCDown(nn.Module):
def __init__(self, c1, c2, k, s):
super().__init__()
self.cv1 = Conv(c1, c2, 1, 1)
self.cv2 = Conv(c2, c2, k=k, s=s, g=c2, act=False)
def forward(self, x):
return self.cv2(self.cv1(x))

67
ultralytics/nn/modules/head.py
View File

@ -12,6 +12,8 @@ from .block import DFL, Proto, ContrastiveHead, BNContrastiveHead
from .conv import Conv from .conv import Conv
from .transformer import MLP, DeformableTransformerDecoder, DeformableTransformerDecoderLayer from .transformer import MLP, DeformableTransformerDecoder, DeformableTransformerDecoderLayer
from .utils import bias_init_with_prob, linear_init from .utils import bias_init_with_prob, linear_init
import copy
from ultralytics.utils import ops
__all__ = "Detect", "Segment", "Pose", "Classify", "OBB", "RTDETRDecoder" __all__ = "Detect", "Segment", "Pose", "Classify", "OBB", "RTDETRDecoder"
@ -40,13 +42,7 @@ class Detect(nn.Module):
self.cv3 = nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch) self.cv3 = nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch)
self.dfl = DFL(self.reg_max) if self.reg_max > 1 else nn.Identity() self.dfl = DFL(self.reg_max) if self.reg_max > 1 else nn.Identity()
def forward(self, x): def inference(self, x):
"""Concatenates and returns predicted bounding boxes and class probabilities."""
for i in range(self.nl):
x[i] = torch.cat((self.cv2[i](x[i]), self.cv3[i](x[i])), 1)
if self.training: # Training path
return x
# Inference path # Inference path
shape = x[0].shape # BCHW shape = x[0].shape # BCHW
x_cat = torch.cat([xi.view(shape[0], self.no, -1) for xi in x], 2) x_cat = torch.cat([xi.view(shape[0], self.no, -1) for xi in x], 2)
@ -74,6 +70,21 @@ class Detect(nn.Module):
y = torch.cat((dbox, cls.sigmoid()), 1) y = torch.cat((dbox, cls.sigmoid()), 1)
return y if self.export else (y, x) return y if self.export else (y, x)
def forward_feat(self, x, cv2, cv3):
y = []
for i in range(self.nl):
y.append(torch.cat((cv2[i](x[i]), cv3[i](x[i])), 1))
return y
def forward(self, x):
"""Concatenates and returns predicted bounding boxes and class probabilities."""
y = self.forward_feat(x, self.cv2, self.cv3)
if self.training:
return y
return self.inference(y)
def bias_init(self): def bias_init(self):
"""Initialize Detect() biases, WARNING: requires stride availability.""" """Initialize Detect() biases, WARNING: requires stride availability."""
m = self # self.model[-1] # Detect() module m = self # self.model[-1] # Detect() module
@ -85,6 +96,8 @@ class Detect(nn.Module):
def decode_bboxes(self, bboxes, anchors): def decode_bboxes(self, bboxes, anchors):
"""Decode bounding boxes.""" """Decode bounding boxes."""
if self.export:
return dist2bbox(bboxes, anchors, xywh=False, dim=1)
return dist2bbox(bboxes, anchors, xywh=True, dim=1) return dist2bbox(bboxes, anchors, xywh=True, dim=1)
@ -480,3 +493,43 @@ class RTDETRDecoder(nn.Module):
xavier_uniform_(self.query_pos_head.layers[1].weight) xavier_uniform_(self.query_pos_head.layers[1].weight)
for layer in self.input_proj: for layer in self.input_proj:
xavier_uniform_(layer[0].weight) xavier_uniform_(layer[0].weight)
class v10Detect(Detect):
max_det = -1
def __init__(self, nc=80, ch=()):
super().__init__(nc, ch)
c3 = max(ch[0], min(self.nc, 100)) # channels
self.cv3 = nn.ModuleList(nn.Sequential(nn.Sequential(Conv(x, x, 3, g=x), Conv(x, c3, 1)), \
nn.Sequential(Conv(c3, c3, 3, g=c3), Conv(c3, c3, 1)), \
nn.Conv2d(c3, self.nc, 1)) for i, x in enumerate(ch))
self.one2one_cv2 = copy.deepcopy(self.cv2)
self.one2one_cv3 = copy.deepcopy(self.cv3)
def forward(self, x):
one2one = self.forward_feat([xi.detach() for xi in x], self.one2one_cv2, self.one2one_cv3)
if not self.export:
one2many = super().forward(x)
if not self.training:
one2one = self.inference(one2one)
if not self.export:
return {"one2many": one2many, "one2one": one2one}
else:
assert(self.max_det != -1)
boxes, scores, labels = ops.v10postprocess(one2one.permute(0, 2, 1), self.max_det)
return torch.cat([boxes, scores.unsqueeze(-1), labels.unsqueeze(-1)], dim=-1)
else:
return {"one2many": one2many, "one2one": one2one}
def bias_init(self):
super().bias_init()
"""Initialize Detect() biases, WARNING: requires stride availability."""
m = self # self.model[-1] # Detect() module
# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1
# ncf = math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum()) # nominal class frequency
for a, b, s in zip(m.one2one_cv2, m.one2one_cv3, m.stride): # from
a[-1].bias.data[:] = 1.0 # box
b[-1].bias.data[: m.nc] = math.log(5 / m.nc / (640 / s) ** 2) # cls (.01 objects, 80 classes, 640 img)

33
ultralytics/nn/tasks.py
View File

@ -49,10 +49,15 @@ from ultralytics.nn.modules import (
CBFuse, CBFuse,
CBLinear, CBLinear,
Silence, Silence,
C2fCIB,
PSA,
SCDown,
RepVGGDW,
v10Detect
) )
from ultralytics.utils import DEFAULT_CFG_DICT, DEFAULT_CFG_KEYS, LOGGER, colorstr, emojis, yaml_load from ultralytics.utils import DEFAULT_CFG_DICT, DEFAULT_CFG_KEYS, LOGGER, colorstr, emojis, yaml_load
from ultralytics.utils.checks import check_requirements, check_suffix, check_yaml from ultralytics.utils.checks import check_requirements, check_suffix, check_yaml
from ultralytics.utils.loss import v8ClassificationLoss, v8DetectionLoss, v8OBBLoss, v8PoseLoss, v8SegmentationLoss from ultralytics.utils.loss import v8ClassificationLoss, v8DetectionLoss, v8OBBLoss, v8PoseLoss, v8SegmentationLoss, v10DetectLoss
from ultralytics.utils.plotting import feature_visualization from ultralytics.utils.plotting import feature_visualization
from ultralytics.utils.torch_utils import ( from ultralytics.utils.torch_utils import (
fuse_conv_and_bn, fuse_conv_and_bn,
@ -191,6 +196,9 @@ class BaseModel(nn.Module):
if isinstance(m, RepConv): if isinstance(m, RepConv):
m.fuse_convs() m.fuse_convs()
m.forward = m.forward_fuse # update forward m.forward = m.forward_fuse # update forward
if isinstance(m, RepVGGDW):
m.fuse()
m.forward = m.forward_fuse
self.info(verbose=verbose) self.info(verbose=verbose)
return self return self
@ -294,6 +302,8 @@ class DetectionModel(BaseModel):
s = 256 # 2x min stride s = 256 # 2x min stride
m.inplace = self.inplace m.inplace = self.inplace
forward = lambda x: self.forward(x)[0] if isinstance(m, (Segment, Pose, OBB)) else self.forward(x) forward = lambda x: self.forward(x)[0] if isinstance(m, (Segment, Pose, OBB)) else self.forward(x)
if isinstance(m, v10Detect):
forward = lambda x: self.forward(x)["one2many"]
m.stride = torch.tensor([s / x.shape[-2] for x in forward(torch.zeros(1, ch, s, s))]) # forward m.stride = torch.tensor([s / x.shape[-2] for x in forward(torch.zeros(1, ch, s, s))]) # forward
self.stride = m.stride self.stride = m.stride
m.bias_init() # only run once m.bias_init() # only run once
@ -627,6 +637,9 @@ class WorldModel(DetectionModel):
return torch.unbind(torch.cat(embeddings, 1), dim=0) return torch.unbind(torch.cat(embeddings, 1), dim=0)
return x return x
class YOLOv10DetectionModel(DetectionModel):
def init_criterion(self):
return v10DetectLoss(self)
class Ensemble(nn.ModuleList): class Ensemble(nn.ModuleList):
"""Ensemble of models.""" """Ensemble of models."""
@ -869,6 +882,9 @@ def parse_model(d, ch, verbose=True): # model_dict, input_channels(3)
DWConvTranspose2d, DWConvTranspose2d,
C3x, C3x,
RepC3, RepC3,
PSA,
SCDown,
C2fCIB
}: }:
c1, c2 = ch[f], args[0] c1, c2 = ch[f], args[0]
if c2 != nc: # if c2 not equal to number of classes (i.e. for Classify() output) if c2 != nc: # if c2 not equal to number of classes (i.e. for Classify() output)
@ -880,7 +896,7 @@ def parse_model(d, ch, verbose=True): # model_dict, input_channels(3)
) # num heads ) # num heads
args = [c1, c2, *args[1:]] args = [c1, c2, *args[1:]]
if m in (BottleneckCSP, C1, C2, C2f, C2fAttn, C3, C3TR, C3Ghost, C3x, RepC3): if m in (BottleneckCSP, C1, C2, C2f, C2fAttn, C3, C3TR, C3Ghost, C3x, RepC3, C2fCIB):
args.insert(2, n) # number of repeats args.insert(2, n) # number of repeats
n = 1 n = 1
elif m is AIFI: elif m is AIFI:
@ -897,7 +913,7 @@ def parse_model(d, ch, verbose=True): # model_dict, input_channels(3)
args = [ch[f]] args = [ch[f]]
elif m is Concat: elif m is Concat:
c2 = sum(ch[x] for x in f) c2 = sum(ch[x] for x in f)
elif m in {Detect, WorldDetect, Segment, Pose, OBB, ImagePoolingAttn}: elif m in {Detect, WorldDetect, Segment, Pose, OBB, ImagePoolingAttn, v10Detect}:
args.append([ch[x] for x in f]) args.append([ch[x] for x in f])
if m is Segment: if m is Segment:
args[2] = make_divisible(min(args[2], max_channels) * width, 8) args[2] = make_divisible(min(args[2], max_channels) * width, 8)
@ -936,7 +952,10 @@ def yaml_model_load(path):
LOGGER.warning(f"WARNING ⚠️ Ultralytics YOLO P6 models now use -p6 suffix. Renaming {path.stem} to {new_stem}.") LOGGER.warning(f"WARNING ⚠️ Ultralytics YOLO P6 models now use -p6 suffix. Renaming {path.stem} to {new_stem}.")
path = path.with_name(new_stem + path.suffix) path = path.with_name(new_stem + path.suffix)
unified_path = re.sub(r"(\d+)([nslmx])(.+)?$", r"\1\3", str(path)) # i.e. yolov8x.yaml -> yolov8.yaml if "v10" not in str(path):
unified_path = re.sub(r"(\d+)([nsblmx])(.+)?$", r"\1\3", str(path)) # i.e. yolov8x.yaml -> yolov8.yaml
else:
unified_path = path
yaml_file = check_yaml(unified_path, hard=False) or check_yaml(path) yaml_file = check_yaml(unified_path, hard=False) or check_yaml(path)
d = yaml_load(yaml_file) # model dict d = yaml_load(yaml_file) # model dict
d["scale"] = guess_model_scale(path) d["scale"] = guess_model_scale(path)
@ -959,7 +978,7 @@ def guess_model_scale(model_path):
with contextlib.suppress(AttributeError): with contextlib.suppress(AttributeError):
import re import re
return re.search(r"yolov\d+([nslmx])", Path(model_path).stem).group(1) # n, s, m, l, or x return re.search(r"yolov\d+([nsblmx])", Path(model_path).stem).group(1) # n, s, m, l, or x
return "" return ""
@ -982,7 +1001,7 @@ def guess_model_task(model):
m = cfg["head"][-1][-2].lower() # output module name m = cfg["head"][-1][-2].lower() # output module name
if m in {"classify", "classifier", "cls", "fc"}: if m in {"classify", "classifier", "cls", "fc"}:
return "classify" return "classify"
if m == "detect": if m == "detect" or m == "v10detect":
return "detect" return "detect"
if m == "segment": if m == "segment":
return "segment" return "segment"
@ -1014,7 +1033,7 @@ def guess_model_task(model):
return "pose" return "pose"
elif isinstance(m, OBB): elif isinstance(m, OBB):
return "obb" return "obb"
elif isinstance(m, (Detect, WorldDetect)): elif isinstance(m, (Detect, WorldDetect, v10Detect)):
return "detect" return "detect"
# Guess from model filename # Guess from model filename

16
ultralytics/utils/loss.py
View File

@ -147,7 +147,7 @@ class KeypointLoss(nn.Module):
class v8DetectionLoss: class v8DetectionLoss:
"""Criterion class for computing training losses.""" """Criterion class for computing training losses."""
def __init__(self, model): # model must be de-paralleled def __init__(self, model, tal_topk=10): # model must be de-paralleled
"""Initializes v8DetectionLoss with the model, defining model-related properties and BCE loss function.""" """Initializes v8DetectionLoss with the model, defining model-related properties and BCE loss function."""
device = next(model.parameters()).device # get model device device = next(model.parameters()).device # get model device
h = model.args # hyperparameters h = model.args # hyperparameters
@ -163,7 +163,7 @@ class v8DetectionLoss:
self.use_dfl = m.reg_max > 1 self.use_dfl = m.reg_max > 1
self.assigner = TaskAlignedAssigner(topk=10, num_classes=self.nc, alpha=0.5, beta=6.0) self.assigner = TaskAlignedAssigner(topk=tal_topk, num_classes=self.nc, alpha=0.5, beta=6.0)
self.bbox_loss = BboxLoss(m.reg_max - 1, use_dfl=self.use_dfl).to(device) self.bbox_loss = BboxLoss(m.reg_max - 1, use_dfl=self.use_dfl).to(device)
self.proj = torch.arange(m.reg_max, dtype=torch.float, device=device) self.proj = torch.arange(m.reg_max, dtype=torch.float, device=device)
@ -713,3 +713,15 @@ class v8OBBLoss(v8DetectionLoss):
b, a, c = pred_dist.shape # batch, anchors, channels b, a, c = pred_dist.shape # batch, anchors, channels
pred_dist = pred_dist.view(b, a, 4, c // 4).softmax(3).matmul(self.proj.type(pred_dist.dtype)) pred_dist = pred_dist.view(b, a, 4, c // 4).softmax(3).matmul(self.proj.type(pred_dist.dtype))
return torch.cat((dist2rbox(pred_dist, pred_angle, anchor_points), pred_angle), dim=-1) return torch.cat((dist2rbox(pred_dist, pred_angle, anchor_points), pred_angle), dim=-1)
class v10DetectLoss:
def __init__(self, model):
self.one2many = v8DetectionLoss(model, tal_topk=10)
self.one2one = v8DetectionLoss(model, tal_topk=1)
def __call__(self, preds, batch):
one2many = preds["one2many"]
loss_one2many = self.one2many(one2many, batch)
one2one = preds["one2one"]
loss_one2one = self.one2one(one2one, batch)
return loss_one2many[0] + loss_one2one[0], torch.cat((loss_one2many[1], loss_one2one[1]))

15
ultralytics/utils/ops.py
View File

@ -847,3 +847,18 @@ def clean_str(s):
(str): a string with special characters replaced by an underscore _ (str): a string with special characters replaced by an underscore _
""" """
return re.sub(pattern="[|@#!¡·$€%&()=?¿^*;:,¨´><+]", repl="_", string=s) return re.sub(pattern="[|@#!¡·$€%&()=?¿^*;:,¨´><+]", repl="_", string=s)
def v10postprocess(preds, max_det, nc=80):
assert(4 + nc == preds.shape[-1])
boxes, scores = preds.split([4, nc], dim=-1)
max_scores = scores.amax(dim=-1)
max_scores, index = torch.topk(max_scores, max_det, axis=-1)
index = index.unsqueeze(-1)
boxes = torch.gather(boxes, dim=1, index=index.repeat(1, 1, boxes.shape[-1]))
scores = torch.gather(scores, dim=1, index=index.repeat(1, 1, scores.shape[-1]))
scores, index = torch.topk(scores.flatten(1), max_det, axis=-1)
labels = index % nc
index = index // nc
boxes = boxes.gather(dim=1, index=index.unsqueeze(-1).repeat(1, 1, boxes.shape[-1]))
return boxes, scores, labels

3
ultralytics/utils/tal.py
View File

@ -308,7 +308,8 @@ def make_anchors(feats, strides, grid_cell_offset=0.5):
def dist2bbox(distance, anchor_points, xywh=True, dim=-1): def dist2bbox(distance, anchor_points, xywh=True, dim=-1):
"""Transform distance(ltrb) to box(xywh or xyxy).""" """Transform distance(ltrb) to box(xywh or xyxy)."""
lt, rb = distance.chunk(2, dim) assert(distance.shape[dim] == 4)
lt, rb = distance.split([2, 2], dim)
x1y1 = anchor_points - lt x1y1 = anchor_points - lt
x2y2 = anchor_points + rb x2y2 = anchor_points + rb
if xywh: if xywh:

9
ultralytics/utils/torch_utils.py
View File

@ -310,10 +310,11 @@ def get_flops(model, imgsz=640):
imgsz = [imgsz, imgsz] # expand if int/float imgsz = [imgsz, imgsz] # expand if int/float
try: try:
# Use stride size for input tensor # Use stride size for input tensor
stride = max(int(model.stride.max()), 32) if hasattr(model, "stride") else 32 # max stride # stride = max(int(model.stride.max()), 32) if hasattr(model, "stride") else 32 # max stride
im = torch.empty((1, p.shape[1], stride, stride), device=p.device) # input image in BCHW format # im = torch.empty((1, p.shape[1], stride, stride), device=p.device) # input image in BCHW format
flops = thop.profile(deepcopy(model), inputs=[im], verbose=False)[0] / 1e9 * 2 # stride GFLOPs # flops = thop.profile(deepcopy(model), inputs=[im], verbose=False)[0] / 1e9 * 2 # stride GFLOPs
return flops * imgsz[0] / stride * imgsz[1] / stride # imgsz GFLOPs # return flops * imgsz[0] / stride * imgsz[1] / stride # imgsz GFLOPs
raise Exception
except Exception: except Exception:
# Use actual image size for input tensor (i.e. required for RTDETR models) # Use actual image size for input tensor (i.e. required for RTDETR models)
im = torch.empty((1, p.shape[1], *imgsz), device=p.device) # input image in BCHW format im = torch.empty((1, p.shape[1], *imgsz), device=p.device) # input image in BCHW format